BRENNEN

Industries / Health / Medical devices / Medical sterilization

Proving it worked

This is the work of showing that a load really was sterilized: tests of the machine, strips and spore tests inside the load, and the records on which a load is released for use.

5

Milestones achieved, from 1963 to 2020.

2

To watch for: announced, not yet confirmed outside.

3

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1963

    Daily test for trapped air

    A prevacuum sterilizer pumps air out of the chamber and load before letting steam in, because trapped air keeps steam from reaching every surface. In 1958 R. Knox and E. J. K. Penikett reported on how an initial vacuum affects the steam sterilization of dressings to Britain's Medical Research Council working party on pressure-steam sterilizers. In March 1963 J. H. Bowie, J. C. Kelsey and G. R. Thompson described in The Lancet "the Bowie and Dick autoclave tape test". In the form the CDC describes today, it is a pack of folded cotton towels with a test sheet in the middle, run at 134 °C for 3.5 minutes in an empty chamber each day before the first load: an even color change means the air was removed, a spot means air stayed in, and the sterilizer should not be used until it is repaired and passes.

    eutils.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.govcdc.gov

    Credit

    J. H. Bowie, J. C. Kelsey and G. R. Thompson; R. Knox and E. J. K. Penikett

  2. 1990

    Spore results in an hour

    A biological indicator is a strip or vial of bacterial spores that are hard to kill, put through a cycle and then incubated: if the spores grow, the cycle failed. The CDC notes that early spore strips needed up to 7 days of incubation and later self-contained vials up to 48 hours. On 3 July 1990 the FDA cleared 3M's Attest Rapid Readout biological indicator, which detects an enzyme of the spores by the fluorescent product it makes. In 1993 W. A. Rutala, M. F. Gergen and D. J. Weber of the University of North Carolina tested its 1291 model against three 24-hour indicators in flash cycles and found that its 60-minute result paralleled theirs. In the same tests two of three chemical indicators, strips that change color with heat and steam, passed every 2-minute exposure, although 40% to 90% of the spores in the conventional indicators survived those exposures.

    LessonA color change shows that conditions were reached, not that microbes died; AAMI's steam guidance treats the spore test as the only monitor that shows the process killed.

    accessdata.fda.goveutils.ncbi.nlm.nih.govsolventum.com

    Credit

    3M; W. A. Rutala, M. F. Gergen and D. J. Weber

  3. 2007

    Indicators must prove resistance

    In the US, biological indicators for hospital sterilizers are class II medical devices that need FDA clearance before sale. The FDA's final guidance of October 2007 asks makers to report each indicator's spore population, its D-value (the time needed to kill 90% of the spores under stated conditions) and its survival and kill times, measured on the finished product in a resistometer, a test vessel that holds the cycle's conditions precisely.

    fda.govfda.gov

    Credit

    FDA's Center for Devices and Radiological Health

  4. 2017

    One in a million, chosen

    The sterility assurance level (SAL) is the probability that a single live microbe remains on an item after sterilization. Europe's standard EN 556-1, also adopted in countries such as Australia and China, lets a terminally sterilized device be called sterile only at an SAL of one in a million (10⁻⁶) or better, while the US standard AAMI ST67 accepts one in a thousand for some products that do not touch broken skin or damaged tissue. The CDC's guideline calls the choice of one in a million "strictly arbitrary", and in 2014 Martell Winters of Nelson Laboratories told MDDI that it was not the result of scientific research. In 2017 ISO published ISO/TS 19930, guidance on choosing a less strict level for products that cannot survive a full sterilization process, such as those with biological components, leaving the decision for each product to its regulator.

    LessonThe industry's best-known number is a convention, and moving off it is decided product by product.

    cdn.standards.iteh.aicdc.govmddionline.com

    Credit

    European Committee for Standardization (CEN); AAMI; ISO

  5. 2020

    Strips that pass failed cycles

    A type 6 chemical indicator is a strip meant to change color only when a given steam cycle's time and temperature have been met. P. R. Laranjeira of Cardinal Health and colleagues at the University of São Paulo reported in 2020 that type 6 indicators showed an "approved" result in steam cycles that had failed, and called for the standards to change. In a 2015 pilot led by Nathan O'Hara of the University of British Columbia, nine hospitals in seven low- and middle-income countries ran 10 chemical integrator tests each: 78% of the sites had acceptable readings on all 10, yet every one of the 90 tests had at least one of exposure time, temperature or pressure outside the target.

    LessonA strip that passes is not proof on its own; the cycle record has to be read as well.

    eutils.ncbi.nlm.nih.goveutils.ncbi.nlm.nih.gov

    Credit

    P. R. Laranjeira, J. A. G. Bronzatti, R. Q. Souza and K. U. Graziano; N. N. O'Hara and colleagues

Watch for

  1. Spore checks in seconds

    On 12 July 2024 the FDA granted Terragene, of Argentina, a De Novo authorization (the route for a new kind of device) for its Bionova Photon BT225, a self-contained steam indicator described in FDA filings as giving a 7-second readout, creating a new device type for indicators that detect spore growth indirectly. On 18 June 2026 the FDA cleared True Indicating's InstaTRU IBI-05, which its maker says shows within 20 seconds whether live spores are present, with sensitivity and specificity above 95% in its own tests.

    Achieved whena laboratory independent of both makers tests these indicators against conventional spore tests in cycles deliberately cut short, and publishes how often the results agree, as the University of North Carolina team did for the 60-minute indicator in 1993.

    accessdata.fda.govaccessdata.fda.gov

    Claimed by

    Terragene (2024) and True Indicating (June 2026)

  2. Machine learning on cycle data

    Parametric release means releasing a load on the record of its measured cycle conditions instead of waiting for spore tests; a 2023 paper by Brian McEvoy and colleagues called its uptake in the sterilization industry slow. A November 2026 review by John Carrig of Boston Scientific and colleagues at the Technological University of the Shannon in Ireland says that the critical settings of vaporized hydrogen peroxide cycles have not been statistically defined and that the sensor data each cycle produces goes unused, and that evidence from other methods shows machine learning can cut the number of test runs and catch abnormal cycles.

    Achieved whena maker publishes a validation in which a model reading cycle data releases loads, and a regulator or notified body accepts it for routine use.

    eutils.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov

    Claimed by

    Carrig, White, Mongan, Ottaviani, Butan and Rowan, Journal of Applied Microbiology, November 2026

Open

  1. 01

    Load records with planted errors

    A public set of made-up sterile processing records for one month in an imaginary hospital: sterilizer cycle printouts, daily air-removal test results, spore and integrator results, implant load logs, tray tracking scans and load recalls. Into them go 100 planted errors of known kinds, such as an implant released before its spore result with no emergency record, loads run after a failed air-removal test, a positive spore test with no recall back to the last negative one, and a cycle shorter than the instrument maker's instructions. An answer key lets anyone score tracking software, AI models and people.

    The brief
    Credit · unclaimed
  2. 02

    Every cleared indicator, tabulated

    An open table of every sterilization indicator the FDA has cleared: maker, 510(k) number and date, sterilization method and cycles, test organism, spore population, D-value, survival and kill times and readout time, and for chemical indicators the colors before and after. Every value carries the page of the 510(k) summary it came from.

    The brief
    Credit · unclaimed
  3. 03

    Sterility recalls, read and coded

    An open, coded table of FDA device recalls whose stated reason mentions sterility: what failed (seal or pouch, sterilization skipped or incomplete, cycle out of specification, contract sterilizer error, failed spore test or dose audit, gap in validation, product wrongly labeled sterile), the sterilization method, and who found the problem, each linked to its recall record.

    The brief
    Credit · unclaimed

Open 01 · the brief

Load records with planted errors

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of made-up sterile processing records for one month in an imaginary hospital: sterilizer cycle printouts, daily air-removal test results, spore and integrator results, implant load logs, tray tracking scans and load recalls. Into them go 100 planted errors of known kinds, such as an implant released before its spore result with no emergency record, loads run after a failed air-removal test, a positive spore test with no recall back to the last negative one, and a cycle shorter than the instrument maker's instructions. An answer key lets anyone score tracking software, AI models and people.

Why it is still open

AAMI's steam guidance says implant loads should be held until the spore result is known, and that a positive result means recalling every load back to the last negative test. In a 2025 survey across ten Asia-Pacific countries, 27% of respondents had seen an implant load released before its spore result, and 21% did not monitor every implant load as WHO and AAMI advise; the authors are paid consultants to Solventum, which sells indicators. Software vendors such as Oxmaint say their systems escalate any failure automatically, but searches found no published test of any system against known errors. Open-source logbooks exist, such as a June 2026 GitHub project that records autoclave cycles and spore tests, but none comes with records that hold known errors. The nearest measured case of AI reviewing manufacturing records is Enveda's drug-record reviewer, measured only against Enveda's own reviewers (see the Note "Seven more flagged").

solventum.comeutils.ncbi.nlm.nih.govgithub.comoxmaint.com

Built with what exists

Python to generate the records from the cycle settings in the CDC guideline and the release rules in AAMI ST79; printouts rendered as PDFs and phone photos; Claude and Gemini as reviewers, each told to cite the rule behind every finding, as the code reviewer in the Note "Findings that cite notes" does; two certified technicians (such as holders of the CRCST certificate) to confirm each planted error.

  1. First step

    Write one week of records for two sterilizers, plant 15 errors, and count how many one model finds.

  2. Done looks like

    At least 2,000 records holding 100 planted errors, each confirmed by two certified technicians who did not plant them; three AI models and two technicians scored on the share of errors found and the share of their flags that are wrong; the generator and scorer public, so anyone can make a fresh month.

Open 02 · the brief

Every cleared indicator, tabulated

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open table of every sterilization indicator the FDA has cleared: maker, 510(k) number and date, sterilization method and cycles, test organism, spore population, D-value, survival and kill times and readout time, and for chemical indicators the colors before and after. Every value carries the page of the 510(k) summary it came from.

Why it is still open

openFDA lists 206 clearances under product code FRC (biological indicators), 266 under JOJ (chemical indicators) and 3 under QVB (indicators with indirect growth detection), but the values sit in PDF summaries, and older clearances may have none online. Sites such as 510kdatabase.net and Innolitics list the clearances by product code without a comparable table of resistance figures or colors, and the side-by-side tables that exist are makers' own, such as Solventum's comparison of its Attest line with Terragene's. The colors matter: in a 2026 study of 902 Chinese health workers, 32.71% reported difficulty reading chemical indicators, and color changes varied across makers and methods.

api.fda.gov510kdatabase.net510k.innolitics.comassets.solventum.comeutils.ncbi.nlm.nih.gov

Built with what exists

the openFDA 510(k) API for the list; the FDA's PDF summaries; the fields the FDA's 2007 guidance asks for as the schema; a model built for pulling fields from long documents, such as Meta's Muse Ember (see the Note "The jobs agents repeat"); DuckDB; GitHub or Zenodo.

fda.gov

  1. First step

    Pull 3M's 35 FRC clearances from 1990 to 2025 and tabulate their readout times and D-values by hand, then compare a model's extraction.

  2. Done looks like

    Every FRC, JOJ and QVB clearance with a summary online tabulated; a second person checks 150 randomly chosen values and finds at least 95% correct; a color table for every chemical indicator whose summary states its colors.

Open 03 · the brief

Sterility recalls, read and coded

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open, coded table of FDA device recalls whose stated reason mentions sterility: what failed (seal or pouch, sterilization skipped or incomplete, cycle out of specification, contract sterilizer error, failed spore test or dose audit, gap in validation, product wrongly labeled sterile), the sterilization method, and who found the problem, each linked to its recall record.

Why it is still open

On 20 November 2026 openFDA held 9,284 recall product entries whose reason mentions sterile or sterility. FDA's own root-cause field puts 1,158 of them under "Other", 1,076 under "Under Investigation by firm" and 201 under unknown. A June 2026 analysis by MedDeviceGuide sorted a 5,000-record sample of all device recalls by keyword and called sterility the most common explicit reason, without coding where the failures happened. The FDA's own 2021 review of 437 Class II recalls ranked root causes by device type, with process control first, but did not single out sterility.

api.fda.govfda.govmeddeviceguide.com

Built with what exists

the openFDA device recall API; a model that scores fixed-choice questions in bulk, such as Gemini Decision 1 (see the Note "A thousand questions, one pass"); two reviewers; DuckDB.

  1. First step

    Download the 2025 entries, code 200 by hand into eight categories, and compare one model's codes with yours.

  2. Done looks like

    Every entry coded and grouped by recall event; two reviewers agree on at least 90% of 300 randomly chosen events; counts by failure point, method and year are published with a script that reruns monthly.